predictive biomarkers for cancer immunotherapy with …review open access predictive biomarkers for...

Bai et al. Biomarker Research (2020) 8:34 https://doi.org/10.1186/s40364-020-00209-0

REVIEW Open Access

Predictive biomarkers for cancer
immunotherapy with immune checkpointinhibitors Rilan Bai, Zheng Lv, Dongsheng Xu and Jiuwei Cui*
Abstract

Although the clinical development of immune checkpoint inhibitors (ICIs) therapy has ushered in a new era of anti-tumor therapy, with sustained responses and significant survival advantages observed in multiple tumors, mostpatients do not benefit. Therefore, more and more attention has been paid to the identification and developmentof predictive biomarkers for the response of ICIs, and more in-depth and comprehensive understanding has beencontinuously explored in recent years. Predictive markers of ICIs efficacy have been gradually explored from theexpression of intermolecular interactions within tumor cells to the expression of various molecules and cells intumor microenvironment, and been extended to the exploration of circulating and host systemic markers. With thedevelopment of high-throughput sequencing and microarray technology, a variety of biomarker strategies havebeen deeply explored and gradually achieved the process from the identification of single marker to thedevelopment of multifactorial synergistic predictive markers. Comprehensive predictive-models developed byintegrating different types of data based on different components of tumor-host interactions is the direction offuture research and will have a profound impact in the field of precision immuno-oncology. In this review, wedeeply analyze the exploration course and research progress of predictive biomarkers as an adjunctive tool totumor immunotherapy in effectively identifying the efficacy of ICIs, and discuss their future directions in achievingprecision immuno-oncology.

Keywords: Neoplasm, Immune checkpoint inhibitor, Predictive biomarker, Tumor mutation burden, Programmeddeath ligand-1

BackgroundImmune checkpoint inhibitors (ICIs) therapy has usheredin a new era of anti-tumor therapy, with sustained re-sponses and significant survival advantages observed inmultiple tumors. Anti-programmed cell death-1/pro-grammed cell death-ligand 1 (PD-1/PD-L1) antibody hasbeen approved for second-line or first-line treatment in avariety of malignant neoplasms, including melanoma, lungcancer, renal cell carcinoma (RCC), head and neck squa-mous cell carcinoma (HNSCC) and gastroesophageal

© The Author(s). 2020 Open Access This articwhich permits use, sharing, adaptation, distribappropriate credit to the original author(s) andchanges were made. The images or other thirlicence, unless indicated otherwise in a creditlicence and your intended use is not permittepermission directly from the copyright holderThe Creative Commons Public Domain Dedicadata made available in this article, unless othe

* Correspondence: [email protected] Center, the First Hospital of Jilin University, 71 Xinmin Street,Changchun, Jilin 130021, China

cancer [1, 2]. However, despite the breakthrough in clin-ical treatment with ICIs, most patients do not benefit.Pembrolizumab or nivolumab has an objective responserate (ORR) of 40–45% in first-line melanoma and 20% insecond-line non-small cell lung cancer (NSCLC) [3–5].Therefore, in recent years, more and more attentions havebeen paid to the identification and development of pre-dictive biomarkers for the efficacy of ICIs, and more in-depth and comprehensive understanding has also beenobtained in recent years, including new data on bio-markers of tumor genome and neoantigen, tumor im-mune microenvironment phenotype, liquid biopsybiomarkers, host-related factors and all of which have

le is licensed under a Creative Commons Attribution 4.0 International License,ution and reproduction in any medium or format, as long as you givethe source, provide a link to the Creative Commons licence, and indicate if

d party material in this article are included in the article's Creative Commonsline to the material. If material is not included in the article's Creative Commonsd by statutory regulation or exceeds the permitted use, you will need to obtain. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.tion waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to therwise stated in a credit line to the data.

http://crossmark.crossref.org/dialog/?doi=10.1186/s40364-020-00209-0&domain=pdf

http://creativecommons.org/licenses/by/4.0/

http://creativecommons.org/publicdomain/zero/1.0/

mailto:[email protected]

Bai et al. Biomarker Research (2020) 8:34 of 17

made many new advances in the corresponding fields.With the development and continuous improvement ofmultiplex immunohistochemical technology, high-throughput sequencing and microarray technology, a var-iety of biomarker strategies have emerged and graduallyrealized the process from the identification of singlemarker to the development of multifactorial synergisticpredictive markers. The development of predictive bio-markers contributes to revealing the therapeutic mecha-nisms of ICIs and the interaction mechanisms betweentumor and host immunity, achieving decision-making ofindividualized anti-tumor immunotherapy, monitoringefficacy and disease development, guiding clinical trial de-sign, as well as for further understanding of drug resist-ance mechanisms and tumor prognosis. In this review, wedeeply analyze the exploration course and research pro-gress of predictive biomarkers as an adjunctive tool totumor immunotherapy in effectively identifying the effi-cacy of ICIs. It should be pointed out here that when read-ing and collating, we try to read and include all therelevant articles. In the process of selecting articles, we in-clude the authoritative articles published in high-leveljournals or the latest research results, and objectively de-scribe and analyze their roles in this field, as well as dis-cuss the reasons that different research results may beinvolved.

Advances of multiple predictive biomarkers toICIs efficacy

(i). Tumor genome and neoantigen biomarkers

Tumor mutation burdenSignificant correlations between high tumor mutationburden (TMB) and response to ICIs have been reported inseveral cancer types [6], including urothelial carcinoma[7], small cell lung cancer (SCLC) [8], NSCLC [9–11],melanoma [12], and human papilloma virus (HPV)-nega-tive HNSCC [13]. A meta-analysis of 27 cancer typesshowed that the mean response rate was positively corre-lated with log (TMB) [14]. The National ComprehensiveCancer Network (NCCN) guidelines have adopted TMBas the recommended test for patients with NSCLC receiv-ing immunotherapy. Although the results in some clinicalstudies of RCC [15], HPV-positive HNSCC [13], and mel-anoma receiving anti-PD-1 after recurrence [16] showedthat TMB alone also did not clearly distinguish respondersand predict OS, it is still exciting that multiple studies inthe 2020 American Society of Clinical Oncology (ASCO)meeting have confirmed the predictive value of TMB inimmunization or combination therapy (KEYNOTE-061study [17, 18], CONDOR study [19], EAGLE study [20],EPOC1704 study [21], etc.), consolidating its status ofTMB as an independent predictor. And in April 2020, the

U.S. Food and Drug Administration (FDA) prioritized theapproval of TMB as a companion diagnostic biomarkerfor pembrolizumab.Nonetheless, the cut-off values of TMB were defined

differently across studies and assay platforms, such asatezolizumab > 16 mt/Mb in urothelial cancer, pembroli-zumab > 23.1 mt/Mb in NSCLC, and atezolizumab≥13.5, ≥15.8, or ≥ 17.1 mt/Mb in NSCLC [22–25], andnivolumab plus ipilimumab ≥10 mt/Mb in NSCLC [10,26], which needs further study to confirm the optimalcut-off value in different tumors. Moreover, the NGSpanels have approved by the FDA that can be used to es-timate TMB include the MSK-IMPACT and Foundatio-nOne CDx panel, the detection results of which arehighly consistent with whole exome sequencing (WES)[11, 27], and other solutions are under development. Astudy detecting TMB (cut-off value at 20 mt/Mb) in4064 NSCLC patients with the FoundationOne platformcontaining a 395 gene panel found that compared withTMB-L patients, overall survival (OS) and DCR was sig-nificantly improved in TMB-H patients treated withanti-PD-1/L1 drug [11]. Both WES and targeted NGS (a422-cancer-gene panel) performed in 78 patients withNSCLC treated with anti-PD-1/L1 demonstrated thatTMB-H population has a significantly better durableclinical benefit (DCB) and progression-free survival(PFS) [27]. These findings demonstrate the feasibility ofcomprehensive genomic profiling (CGP), but the designof optimal next generation sequencing (NGS) panel thatis more accurate, comprehensive and cost-effective isstill not clear. In addition, given that bTMB was identi-fied as a predictor of PFS but failed to differentiate pa-tients with OS benefits, researchers consider the need toexplore other more precise factors, e.g. allele frequency(AF). A study that developed a new bTMB algorithm intwo independent cohorts (POPLAR and OAK) showedthat modified bTMB, low AF bTMB (LAF-bTMB, muta-tion counts with an AF < 5%), was significantly associ-ated with favorable (HR = 0.70, 95%CI 0.52–0.95, p =0.02), PFS (HR = 0.62, 95%CI 0.47–0.80, p < 0.001), andORR (p < 0.001) after immunotherapy, but required tobe prospectively validated [28]. Finally, static biomarkersare insufficient to accurately predict response due to thecomplexity of tumor-immune interactions. A recent ana-lysis of tumor genome-wide dynamic detection in pre-treatment and on-treatment melanomas found thatpretreatment TMB was only associated with OS in un-treated patients, while early (4-week) on-treatment changein TMB (ΔTMB) was strongly associated with anti-PD-1response and OS in the entire cohort [16]. The detectionof ΔTMB is helpful for early evaluating the response totherapy of patient, but its clinical usability limited by thedifficulty in obtaining tissue samples and high price, whileliquid biopsy discussed below might better.


In addition, epigenetic changes are associated withTMB. The latest study investigated the association be-tween TMB and DNA methylation (DNAm) to explorepotential complimentary biomarkers for NSCLC im-munotherapies. The results showed that high TMBNSCLCs had more DNAm aberrance and copy numbervariations (CNVs), showing certain value in predictingefficacy, such as HOX gene methylation status and TMB[29] Thus, the correlated exploration of epigenetics hasattracted more attention in recent years, and liquidbiopsy-based epigenetic studies may become a future re-search direction. Exploration in Chinese NSCLC patientsshowed that NSCLCs with high TMB had DNAm aber-rance and CNVs. Some insertion and deletion (indel)mutations can lead to frameshifts and more immuno-genic neoantigens [30]. In the pan-cancer analysis of 19cancer types evaluated in The Cancer Genome Atlas(TCGA), RCC had the highest indel mutation load, andframeshift indel mutations were found to produce threetimes more candidate neoantigens per mutation thannonsynonymous single nucleotide variants (nsSNVs)[30]. Somatic copy number alterations (SCNAs) are an-other feature of the genomic landscape of tumors, andpan-cancer TCGA analysis revealed an inverse correl-ation between SCNAs at the single-arm or wholechromosome-level and immune infiltration in 10 tumortypes tested [31], and this result was subsequently repli-cated in a larger study of TCGA [32].

DNA damage response pathwaysGenetic variation involved in DNA mismatch repair(MMR) pathway can lead to microsatellite instability(MSI), a specific type of high TMB tumors, and in-creased numbers of CD8+ tumor infiltrating lymphocytes(TILs), PD-1+TILs, and indoleamine 2,3-dioxygenase(IDO)+ tumor cells have been shown in MMR deficiency(dMMR) colorectal cancer [33]. Recently, five clinical tri-als (Keynote-016, 164, 012, 028, 158) including multipletumor types have shown that patients with dMMR/MSI-H can achieve durable responses to pembrolizmab.Based on this, pembrolizumab is approved by the U.S.FDA for the treatment of any advanced solid tumor withdMMR/MSI-H, and nivolumab in combination with ipi-limumab has also shown promising response in dMMR/MSI-H colorectal cancer [34]. In addition, dMMR canalso cause mutations in the DNA polymerase gene epsi-lon/delta 1 (POLE/POLD1), increasing the mutationload and neoantigen load. Analysis of POLE/POLD1mutations in 47,721 patients with different cancer typesshowed that patients with these mutations had signifi-cantly higher TMB and OS. Therefore, it may be an in-dependent risk factor and prognostic marker foridentifying patients who benefit from ICIs [35]. Inaddition, pathways of base excision repair (BER),

homologous recombination repair (HRR), MMR in theDNA damage response (DDR) signaling network con-tribute more significantly to TMB or neoantigens, whichhave the highest levels when co-mutated [36]. It hadbeen identified that co-mutations in the DDR pathwaysof HRR and MMR or HRR and BER, defined as co-mut+,are associated with increased levels of TMB, neoantigenload and immune gene expression signatures. Co-mut+

patients showed a higher ORR and longer PFS or OS, in-dicating that co-mut can be used as predictors of re-sponse to ICIs and provide a potentially convenientmethod for future clinical practice [36].

Specific mutated gene pathways in tumor cellsIt is worth noting that alterations of signaling pathwaysin tumor cells affect the responsiveness to immunother-apy. Patients with mutations in the interferon (IFN)-γpathway genes, IFNGR1/2, JAK1/2, and IRF1, are poorlyresponsive to ICIs treatment and confer resistance [37].A study found that in patients receiving immunotherapy,tumor cells can downregulate or alter IFN-γ signalingpathways such as loss-of-function alleles of genes encod-ing for JAK1/2, and changes in STAT1 to escape the in-fluence of IFN-γ [38], resulting in poor efficacy andresistance. Recent studies suggest that inactivating muta-tions in a mammalian analog of the chromatin remodel-ing SWI/SNF complex and unique genes of the PBAFcomplex (Pbrm1, Arid2, and Brd7) lead to sensitivitiesto ICIs [39, 40]. Loss of function of the PBAF complexincreased chromatin accessibility to transcription regula-tor elements of IFN-γ–inducible genes within tumorcells, and subsequently increased production of CXCL9/CXCL10 chemokines, leading to more efficient recruit-ment of effector T cells into tumors [41]. In human can-cers, expression of Arid2 and Pbrm1 are related toexpression of T cell cytotoxicity genes, which confirmedin Pbrm1-deficient murine melanomas with strongly in-filtrated by cytotoxic T cells and responsive to immuno-therapy [15, 41]. In addition, double-stranded RNA(dsRNA) editing enzyme adenosine deaminase acting onRNA (ADAR1) protein can block the IFN-γ signalingpathway and lead to poor ICIs efficacy and resistance.Loss of function of ADAR1 in tumor cells can reduce A-to-I editing of interferon-inducible RNA species and leadto dsRNA ligand sensing by PKR and melanomadifferentiation-associated protein 5 (MDA5). This resultsin growth inhibition and tumor inflammation, respect-ively, and profoundly sensitizes tumors to immunother-apy [42]. Finally, demethylation positively regulates thetranscriptional activity of some immune-related genes,including PD-L1 and IFN signaling pathway genes, sensi-tizing it to anti-cytotoxic T-lymphocyte-associatedprotein-4 (CTLA-4) therapy [43].


In addition to the IFN-γ-related signaling pathway, al-terations in other tumor genome, such as tumor onco-genes and suppressor genes pathways, and pathwaysrelated to tumor cell proliferation and infiltration, canalso affect immunotherapy efficacy. Epidermal growthfactor receptor (EGFR) and anaplastic lymphoma kinase(ALK) mutations have been shown to be associated withreduced response rates to ICIs and low TMB, and there-fore the FDA does not recommend first-line ICIs-treatment in patients with EGRF or ALK positive tumors[44, 45]; certain types of mutations in MDM2/MDM4and ARID1A can predict non-response to ICIs in highTMB tumors [46]; NSCLC with KRAS and STK11 co-mutated was associated with reduced response andshorter survival in three independent cohorts of patientstreated with anti-PD-1 therapy [47], and STK11 defi-ciency was an independent indicator of poor anti-PD-1response in NSCLC with KRAS mutant; however, at the2020 American Association for Cancer Research (AACR)meeting, 33.7% of patients in the Keynote-042 study(NCT02220894) update data were tested for STK11 andKEAP1, and the results showed that patients could bene-fit from pembrolizumab regardless of STK11 and KEAP1status, but patients with STK11 mutations did not re-spond well to chemotherapy, but given that only 1/3 ofall patients had mutation detection, the results may beaffected; in initial data from studies using targeted NGSpanels suggested that duration of ICIs-treatment was as-sociated with certain BRAF and MET alterations, butnot TMB status [48]. NOTCH signaling pathway is asso-ciated with the occurrence, development and prognosisof tumors, especially with the biological function of can-cer stem cells. Recent breakthrough findings have distin-guished deleterious NOTCH mutation, showing that itcan be used as a potential predictor of favorable ICI re-sponse in NSCLC, potentially via greater transcription ofgenes related to DNA damage response and immune ac-tivation [49]. Another tumor-specific inheritance thatmay influence ICIs efficacy is the aberrant expression ofendogenous retroviruses (ERVs). Pan-cancer analysisidentified a positive correlation of transcript expressionof ERVs with T-cell activity in various tumors [50] andpatient prognosis [51]. Furthermore, with the improve-ment of precision detection technology, the accurateanalysis of negative mutation sites helps to identify thepossibly effective ones. For example, the analysis of studydata of second-line PD-1/L1 inhibitor therapy found thatthe mPFS of patients with KRAS G12C or G12V was sig-nificantly better than that of patients with KRAS muta-tions at other sites [52].In addition, several pan-cancer biomarkers are recently

approved by the FDA. For example, given the effectiveORR of 35.5% and a disease control rate (DCR) of 82%in second-line cholangiocarcinoma patients treated with

pemigatinib, a new targeted therapy, the recent FDA ap-proval of pemigatinib for the treatment of previouslytreated patients with locally advanced or metastatic chol-angiocarcinoma with fibroblast growth factor receptor 2(FGFR2) fusion or rearrangement, and the comprehen-sive genomic analysis assay, FoundationOne CDx,developed by Foundation Medicine as a companiondiagnostic. Also exciting is the recent FDA approval ofthe targeted anticancer drug capmatinib for the treat-ment of metastatic NSCLC with MET exon 14 skipping(METex14) mutations, including first-line patients andpreviously treated patients, also using FoundationOneCDx as a companion diagnostic to help detect specificmutations present in tumor tissue.

Neoantigen loadNeoantigen load, the number of mutations actually tar-geted by T cells, may be directly related to the responseto ICIs [53–55]. A retrospective study showed thatclonal neoantigen burden was associated with the longerOS in primary lung adenocarcinomas (p = 0.025) [53].Traditionally, computational neoantigen predictionshave focused on major histocompatibility complex(MHC) binding of peptides based on anchor residueidentities, however, neoantigen loads identified by thismethod are generally not superior to overall TMB inpredicting ICIs efficacy or survival [56]. In recent prac-tice, this neoantigen can be assessed by the difference inpredicted MHC-I binding affinity between the wild-typepeptide and the corresponding mutant peptide, knownas the differential agretopicity index (DAI), reflectingclinically relevant immunogenicity of tumor peptide[57]. A high DAI value indicates that the mutant peptidesignificantly increases binding affinity to MHC com-pared to the wild-type sequence and can generate moreimmune responses. Studies on previously published co-horts treated with three ICIs have shown that DAI out-performs TMB and the traditionally defined neoantigenload in predicting survival [58, 59]. In addition, lowneoantigen intratumour heterogeneity might also be im-portant for ICIs response. Analysis of the lung adenocar-cinoma TCGA database found that combining highmutational load and low intratumoral neoantigen het-erogeneity (< 1%) was significantly associated with OSand longer lasting clinical benefit than either variablealone [53]. Another reported method for assessingneoantigen foreignness is based on sequence homologyof experimentally validated immunogenic microbial epi-topes in the Immune Epitope Database (IEDB) [60], butit does not account for all possible human leukocyteantigen (HLA) contexts. In addition, the detection forneoantigen can be reflected from different levels such aspeptides or genomes. A study developed the Neopepseealgorithm using a machine learning approach incorporating


integration of nine immunogenicity features and gene mu-tation expression levels [61], and its application to melan-oma and leukemia patients could improve the sensitivityand specificity of neoantigen prediction. Recently it has alsobeen shown that promoter hypermethylation of neoantigengenes may be an important mechanism for immune editingand tumor immune evasion [62], indicating that combineddetection of tumor genome and epigenetics may providemore information for immunotherapy efficacy.

(ii).Tumor immune microenvironment phenotypebiomarkers

PD-L1 expressionGiven that multiple studies in a variety of tumors havedemonstrated a positive correlation between PD-L1 ex-pression and response to ICIs or OS, even in first-linecombination therapy [63–65], pembrolizumab is cur-rently approved by the FDA for use in patients with PD-L1+ (PD-L1 ≥ 50% of tumor cells in first-line treatmentand ≥ 1% in second-line treatment) NSCLC and PD-L1immunohistochemistry (IHC) as a companion diagnosticfor anti-PD-1 therapy in NSCLC patients [66, 67]. How-ever, some studies have not detected a significant correl-ation between PD-L1 expression and response to ICIs[5, 13, 68], and PD-L1 negative patients can still benefitclinically with treatment with ICI or combination treat-ment with ICIs [69], with ORRs ranging from 11 to 20%.Therefore, PD-L1 cannot yet be a comprehensive and in-dependent biomarker in clinical practice in assessing ef-ficacy, with following challenges still existing. Firstly,PD-L1 assay and antibody are not standardized [70].Secondly, PD-L1 expression is temporally and spatiallyheterogeneous [71]. A study of 398 metastatic NSCLCtreated with ICIs showed that PD-L1 varies substantiallyacross different anatomic sites and during clinicalcourse, being highest in adrenal, liver and lymph nodemetastases and lower in bone and brain metastases. Andthe predictive value of PD-L1 at different biopsy sites forthe benefit of ICIs in NSCLC may vary: higher PD-L1 inlung or distant metastasis specimens was significantly as-sociated with higher response rate, PFS and OS, whilePD-L1 in lymph node metastasis biopsy was not associ-ated with either response or survival [72]. Thirdly, posi-tive score and cut-off value of PD-L1 expression is notstandardized [71]. At present, PD-L1 positive scoremainly focuses on the PD-L1 expression level of tumorcells, that is, tumor proportion score (TPS). But PD-L1is also expressed on immune cells such as lymphocytesand macrophages and stromal cells, thus the investiga-tors introduce the concept of combined positive score(CPS), which is the proportion score of the sum of PD-L1 expressed by tumor cells and tumor-associated im-mune cells. In addition, PD-L1 expression on immune

cells is also considered separately as one of the biomarkersto distinguish the benefit population, called immune posi-tive score (IPS). Herbst et al. [73] showed that response toatezolizumab treatment was significantly associated withhigh levels of PD-L1 expression on the surface of TILs be-fore treatment, but not with PD-L1 expression on tumorcells (p = 0.079). Finally, other inhibitory immune path-ways may affect the response to ICIs therapy, including Tcell immunoglobulin-3 (TIM-3), lymphocyte activationgene-3 (LAG-3), and V-domain Ig suppressor of T-cell ac-tivation (VISTA), which can be used as potential bio-markers for ICIs response.

Biomarkers of tumor-infiltrating immune cellsOverall immune status of tumor microenvironmentThe pattern of tumor immune infiltration can be broadlyclassified into immune-inflamed, immune-excluded andimmune-desert [74]. Immune-inflamed is characterizedby the presence of CD8+ and CD4+ T cells in the tumorparenchyma accompanied by the expression of immunecheckpoint molecules [75], indicating a potential anti-tumor immune response to ICIs treatment [73];immune-excluded is characterized by the presence ofdifferent immune cell types in the aggressive margin orstroma of tumor, but cannot infiltration into tumor par-enchyma [74, 76]. Analysis of pre-treatment samples foranti-PD-1/PD-L1 revealed a relatively high abundance ofCD8+T cells at the invasive margin in responders, andserial sampling during treatment showed an increasedinfiltration of CD8+T cells into tumor parenchyma [77];while immune-desert phenotype is characterized by theabsence of abundant T cells in the parenchyma orstroma of tumors and poor response to ICI-treatment[73]. Recently, immunoscore has been proposed as avalid marker for characterizing the immune status oftumor microenvironment (TME), classifying tumors, aswell as predicting treatment response and prognosis[78], which involves the density of two lymphocyte pop-ulations (CD8+ and memory [CD45RO+] T cells) in thecenter and invading margin of tumor [79]. Mlecnik et al.[80] evaluated immunoscore in 599 specimens of stageI–IV colorectal tumor and confirmed that it was signifi-cantly associated with PFS, DFS, and OS, and multivari-ate analysis also showed the superiority of immunoscorein predicting disease recurrence and survival. The valueof immunoscore to predicting ICIs efficacy is being vali-dated internationally in clinical trials of melanoma andNSCLC [78].A wider assessment of active immune responses within

TME by immune gene expression profiling might effect-ively predict clinical benefit to ICIs strategies. Analysisof total RNA and genes that were substantially differentbetween the patient groups in 50 pretreatment tumor bi-opsies revealed at least a 2.5-fold increase in the


expression of 22 immune-related genes in clinically ac-tive patients, including cytotoxic T cell markers (e.g.,CD8A, perforin 1, granzyme B), Th1 cytokines or che-mokines, MHC-II, and other immune-related genes (e.g.,NKG7, IDO1) [81]. Ascierto et al. [82] screened morethan 299 immune-related genes in patients with recur-rent breast cancer 1–5 years after treatment and thosewithout recurrence more than 7 years later, and foundthat five genes (IGK, GBP1, STAT1, IGLL5, and OCLN)were highly overexpressed in patients with recurrence-free survival. In addition, IFN-γ-induced immune genesignatures may be effective biomarkers for predicting theclinical benefit of treatment with ICIs. The study devel-oped IFN-γ scores combining multiple immune variablesbased on 10 gene signatures, which were then extendedto 28 gene signatures in a validation set of 62 melanomapatients, including genes encoding IFN-γ, granzymes A/B, perforin 1, IDO1, and other immune-related genes.Both gene scores showed significant associations withbest overall response rate and PFS. Optimized cut-offvalues for IFN-γ scores based on receiver operatingcharacteristic curve (ROC curve) can achieve a positivepredictive value of 59% for responders and a negativepredictive value of 90% for non-responders [83].

Immune cells with specific phenotypes in TMEThe phenotype of TILs also influences the efficacy ofICIs. The study used single-cell mRNA sequencing(scRNA-seq) data analysis to identify two major CD8+Tcell phenotypes within melanoma: memory-like andexhausted [84], the proportion of which is strongly cor-related with response to ICIs. The research furtherfound that the transcription factor 7 (TCF7) is selectivelyexpressed in memory-like T cells, so the ratio ofCD8+TCF7+ to CD8+TCF7-TILs is strongly correlatedwith improved response and survival in melanoma pa-tients treated with anti-PD-1 [84]. Balatoni et al. [85]found that 7 of 11 immune cells in TME were positivelyassociated with OS after treatment, including CD4+ andCD8+ T cells, FOXP3+ T cells, CD20+ B cells, CD134+

and CD137+ cells, and NKp46+ cells, and different im-mune cells at different sites were differently associatedwith clinical outcomes. Researchers found that only asmall proportion of CD8+ TILs in tumors couldrecognize tumor mutation-associated antigens, while an-other population (bystander cells) was insensitive, anddifferential CD39 expression was the key molecule thatdistinguished the two populations [86]. Analysis of per-ipheral blood from a patient with colorectal cancer whoresponded rapidly to pembrolizumab treatment showedhigh expression of CD39 on CD8+ TILs, indicating thatCD39+CD8+TIL may be a promising predictive bio-marker [86]. The fact of very low level of CD39 expres-sion on CD8+TILs in 50% of EGFR-mutant NSCLC is

consistent with their low response rate to anti-PD-1immunotherapy.In addition, a study showed that Fc domain glycan of

the drug and Fcγ receptor (FcγR) expressed by the hostbone marrow cells could determine the ability of PD-1-tumor-associated macrophages (TAMs) to capture anti-PD-1 drugs from the surface of T cells, which leads toPD-1 inhibitor resistance [87], and the association ofTAMs and poor anti-PD-1 response was reported inmelanoma cohorts [88]; anti-PD-1 response was associ-ated with an increase in CD8+T cells and natural killercells (NK cells) and a decrease in macrophages [16]; andhigh intratumoral myeloid markers were associated witha nearly 6-fold decrease in mPFS after anti-PD-L1 ther-apy in RCC, emphasizing the inhibitory role of myeloidcells in response to ICIs [89]. In conclusion, immunecells in TME show a great promise in the developmentof predictive biomarkers for ICIs.

Diversity of immune repertoires in TMEEffective T cell responses involve the activation and ex-pansion of specific antigen-reactive T cell clones, so di-versity of immune repertoire in intratumoral orperipheral may correlate with ICIs responses and can bequantified as richness and clonality [16]. However, theresults seem to be complex, with some studies finding apositive correlation between TIL clonality and the re-sponse to ICIs before [90] or after [91] treatment, whileothers showing that only an increase in TIL clonalityduring treatment is associated with the response to anti-PD-1 [16, 92]; others show that intratumoral T cellclonality is not associated with survival, while peripheralT cell clonality is inversely associated with PFS and OS[93]. Tumeh et al. [77] further investigated whetherbaseline TILs have a narrow T cell receptor (TCR) rep-ertoire, focusing on tumor-specific immune responsesand whether this narrow TCR repertoire correlates withpembrolizumab responses. They found that respondingpatient had more restricted usage of the TCR beta chain(ie, a more clonal, less diverse population) than patientswith progressive disease, and showed a 10-times increasein these clones after treatment, implying a tumor-specific response to treatment in these patients. Notably,baseline TCR clonality was not highly correlated withTIL density, suggesting that some patients with re-stricted TCR clonality specific for tumor antigens maystill benefit from anti-PD-1 therapy even though TILdensity is low. Recently, researchers have proposed theimmune repertoire (IR)-Index, the average frequency ofshared TCR clones in T clones in TILs and peripheralPD-1+CD8+ T cells. They found that neoantigen-stimulated TCR agreed with IR-Index, and patients withhigh IR-index had better immune activation and highergene expression profiles (GEPs) score, subsequently they


confirmed the predictive value of IR-index to ICIs effi-cacy (DCR/PFS). But considering that it is difficult tosort out PD-1+CD8+ T cells in tumor tissue, based ontwo separate patient cohorts, a research confirmed thatTCR repertoire diversity and clonality of peripheral PD-1+CD8+T cells may serve as noninvasive predictors ofclinical outcomes after ICIs in patients with NSCLC[94]. The viewpoints of T cell diversity and TCR clonal-ity as markers of ICIs efficacy need to be further vali-dated in a large patient population.

(iii).Liquid biopsy biomarkers

Peripheral blood cell biomarkersPeripheral blood is a non-invasive source to explore po-tential biomarkers for ICIs, and although associationswith clinical benefit and survival have been observed, itseffectiveness has not been validated in prospective stud-ies. Analysis of melanoma treated with ipilimumabshowed that improved OS and PFS were associated withbaseline values of peripheral blood components, includ-ing low absolute neutrophil count, low neutrophil-to-lymphocyte ratio (NLR), low absolute monocyte count,low frequency of myelogenous suppressor cells, high fre-quency of FoxP3+ Treg cells, high lymphocyte frequency,high eosinophil count; and clinical benefit also associ-ated with the dynamic changes of blood markers duringtreatment, including decreased FoxP3+Treg concentra-tions and increased lymphocyte and eosinophil counts[95]. Reports in patients with melanoma treated withpembrolizumab and in patients with NSCLC treatedwith nivolumab have shown that NLR is associated withworse tumor response [96, 97]. Multivariate analysis inmelanoma patients treated with anti-PD-1 antibodiesshowed that NLR was the only factor associated withworse ORR and shorter PFS, indicating that NLR is astrong predictor of worse outcome in patients treatedwith ICI [96]. Low baseline lactate dehydrogenase (LDH)levels, high relative/absolute eosinophil counts, and rela-tive lymphocyte counts were associated with prolongedOS in anti-PD-1 and CTLA-4 treated melanoma [97,98]. Given that previous studies have proposed the im-portance of baseline derived NLR (dNLR) and LDHlevels as prognostic markers, a recent study proposed acomposite prognostic index that comprehensively takesthe two factors into account, lung immune prognosticindex (LIPI), which characterized 3 risk groups: good,intermediate, and poor [99]. The analysis of 3987 pa-tients with advanced NSCLC in 11 randomized trialsshowed that patients with good LIPI score who receivedICI were associated with significantly better PFS and OScompared with patients with poor LIPI score, which wasnot observed in patients received chemotherapy [99].

The study of melanoma treated with ipilimumabshowed that the percentage of baseline CD45RO+/CD8+T cells was ≤25% in 80% of non-responders and ≥30% in all responders (p < 0.01) [100]. CyTOF analysis ofmelanoma treated with ICIs showed that the abundanceof CD69+ and MIP1β+ NK cells [101] andCD14+CD16−HLA-DRhi cells [102] were predictive bio-markers of response to anti-PD-1 therapy. In addition,ipilimumab treatment of melanoma with baseline highlevels of circulating Tregs was associated with OS, pos-sibly as a target for ipilimumab antibody ADCC due toits high CTLA-4 expression; whereas decreased or stabi-lized circulating Tregs at 12 weeks since ipilimumab ini-tial administration was significantly associated withbetter DCR and OS [98]. Inducible T cell co-stimulator(ICOS) is costimulatory molecule expressed by activatedT cells and Tregs. Analysis of surgical tissues and per-ipheral blood before and after treatment showed thatanti-CTLA-4 treatment could induce ICOS pathway ac-tivation, and CD4+ICOS+T cells could produce IFN-γand recognize tumor antigens [103]. In addition, a recentreport correlated the detection of circulating tumor cells(CTCs) in peripheral blood with the metastatic processof tumors, and PD-L1 is highly expressed in CTCs frompatients with advanced head and neck cancer, suggestingthat PD-L1+CTC may be a predictive biomarker of re-sponse to ICIs [104].

Biomarkers of circulating tumor DNAThe detection of circulating tumor DNA (ctDNA) canobtain tumor genomic information related to the re-sponse to ICIs, although the sensitivity or specificity hasyet to be improved. Multiple studies showed that highmutation number of ctDNA was associated with im-proved OS and poor prognosis in patients with differentcancer types treated with ICIs [105, 106]; Lee et al. [107]demonstrated that melanoma patients with persistentlyelevated ctDNA during PD-1 antibody therapy showedworse response and shorter PFS and OS. In addition,ctDNA can be a useful marker for identifying pseudo-progression during ICIs treatment. 9 patients with mel-anoma appeared pseudoprogression after ICIs therapywere reported to have favorable ctDNA profiles (definedas undetectable ctDNA or detectable ctDNA at baselinefollowed by > 10-times decrease in ctDNA), while 18 of20 patients with true progression had unfavorablectDNA profiles [108]. The association of bTMB levelbased on ctDNA and clinical benefit with anti-PD-1/L1therapy was validated in tumor patients, confirming thatit is a promising predictive biomarker. NCC-GP150established using optimized gene panel size and algo-rithms is feasible for bTMB evaluation and bTMB canbe used as a biomarker of clinical benefit in NSCLC pa-tients treated with ICIs [109]. Another similar study


showed that tTMB are strongly correlated (Spearman0.6, Pearson 0.7) with bTMB evaluated with a 500-genepanel, which may serve as a potential biomarker for theefficacy of single and dual immunotherapy when cut-offvalue at 20 mt/Mb [110]. In addition, dynamic monitor-ing of ctDNA can provide ΔbTMB information to pre-dict the responsiveness in the treatment process in anon-invasive manner, potentially improving the sensitiv-ity and specificity of response prediction.

Other circulating molecular biomarkersExosomes in the plasma can also provide informationabout the tumor and immunotherapy. Lower baselinelevels and increases during treatment in circulating exo-somal PD-L1 in melanoma patients were associated withresponse to pembrolizumab [111]. However, in anotherstudy of melanoma or NSCLS treated with anti-PD-1,the expression of PD-1 mRNA in the exosomes washigher at baseline and significantly decreased after treat-ment in patients with response, while it was stable in pa-tients with stable disease and increased in patients withprogressive disease after treatment [112]. Therefore, pro-tein and transcripts of exosomal PD-L1 may providecontradictory information on the response to ICIs andrequire large-scale prospective studies for validation. Inaddition, RNA sequencing analysis of PD-L1 inhibitor-resistant NSCLC patients revealed the presence of PD-L1 variant fragments (v242 and v229, which retain thePD-1 binding domain) in vivo and in peripheral bloodand pleural effusion, resistant patients with variant hadmuch higher sPD-L1 concentrations. Experimentsin vitro and vivo have confirmed the inhibitory effect ofPD-L1 variant fragments on T cell activity [113], indicat-ing a poor efficacy response.In addition, other potential predictive biomarkers for

ICIs efficacy have been preliminarily explored [114], in-cluding soluble proteins (e.g., sCD163, sNKG2DLs), cy-tokines and inflammatory factors [e.g., tumor necrosisfactor (TNF)-α, interleukin (IL)-6, C-reactive protein(CRP)]. Baseline serum LDH is often an independentfactor for poor prognosis and shorter OS with ipilimu-mab or pembrolizumab in patients with advanced mel-anoma [97, 115]. Several studies showed that in patientswith various cancers treated with ICIs, high baselineLDH was associated with poor anti-tumor response[116, 117]. Weber et al. [118] analyzed baseline levels ofCRP and IL-6 in serum from patients with melanomawho participated in 3 different clinical trials and levelsabove baseline median were found to be significantly as-sociated with poor response and shorter survival afternivolumab treatment, and similar results were foundwith ipilimumab and combination therapy. In vitro stud-ies revealed that purified CRP significantly inhibited Tcell activation and proliferation at concentrations >

10 μg/mL [118]. Studies have also demonstrated that IL-6 has an immunosuppressive effect under certain condi-tions, including induction of myeloid-derived suppressorcells (MDSCs), which may explain the above phenomenon[119]. Besides, two retrospective studies involving approxi-mately 3000 patients found that high baseline levels ofplasma IL-8 were significantly associated with poor prog-nosis with PD-1/L1 inhibitors therapy and may be a driverof resistance to ICIs [120, 121]. scRNA-seq of the immunecompartment showed that IL-8 is primarily expressed incirculating and intratumoral myeloid cells, and had an in-hibitory effect on adaptive immunity. High IL-8 levelswere associated with higher tumor neutrophil/monocyteinfiltration, poorer antitumor activity of effector T cells, aswell as weaker antigen presentation. Patients with both ahigher T cell effect profile score and lower plasma IL-8levels can obtain the greatest benefit from ICIs therapy.

(iv).Host-related markers

General characteristicsStudies have shown that gender differences are associ-ated with the responsiveness to anti-tumor immune. Ameta-analysis including 20 randomized controlled trials(RCTs) of ICIs (n = 11,351) reported that gender differ-ence in the efficacy ICIs was significant (p = 0.0019),with pooled OS hazard ratio being 0.72(95%CI 0.65–0.79) in male patients and 0.86(95%CI 0.79–0.93) in fe-male patients [122]. In another meta-analysis of a largenumber of melanoma and NSCLC patients treated withICIs, both PFS and OS were significantly longer in malepatients than in female patients, and this difference wasmore pronounced in melanoma patients and anti-CTLA-4 antibodies [123]. Aging is associated with re-stricted immune function with significant effects on bothinnate and adaptive immune responses [124]. A preclin-ical study showed that aged mice had significantly in-creased tumor responses to anti-PD-1 agents comparedwith young mice, considered to be associated with alower proportion of Tregs in aged mice [125]. Consist-ently, melanoma patients over 60 years old have a signifi-cantly higher tumor response to pembrolizumab thanpatients under 60, and the likelihood of response in-creases with age [125]. However, different results havealso been reported by Nishijima et al. with an associationbetween age less than 75 years and better ORR in pa-tients treated with ICIs [126]. The Checkmate-171 trialshowed that patients ≥70 years of age had comparabletolerability and efficacy to the overall population [127].However, at present, the inclusion and representative-ness of the elderly in clinical studies are still insufficient.Besides, studies of the effect of performance status (PS)

on the efficacy of ICIs have shown that good PS are asso-ciated with lower tumor burden and a predominance of


immune cell function in TME. In the Checkmate-171, pa-tients with PS =2 had inferior efficacy to the overall popu-lation [127], and real-world data in Israel again suggestedthat patients with PS ≥2 had inferior efficacy to the overallpopulation [128]. A study reported by ASCO in 2018 (Jus-tin F. et.al. 2018 ASCO abstract#9011) showed that theTMB of smoking patients with NSCLC in the two groupswith equivalent PD-L1 expression was higher, and PFSand duration of response (DOR) in smoking patients withhigh PD-L1 expression (TPS ≥50%) were longer. Differ-ences in body fat distribution also affects tumor prognosisand immunotherapy [129]. A study showed obesity in-duces T cell exhaustion and dysfunction by affecting PD-1expression through STAT3 signaling, leading to increasedimmune aging and promoting tumor growth and progres-sion [130]. However, targeting receptors on activated Tcells or chimeric antigen receptor (CAR)-T cells in im-munotherapy may help enhance T cell function, especiallyin the presence of high leptin. Studies have found thatobese mice also showed a significantly better response toanti-PD-1 without significant toxic side effects [130],which was reproduced in multiple cancer populations re-ceiving ICIs [131], with higher body mass index (BMI)(BMI > 30) patients showing reduction in tumor burdenand improvement in PFS and OS. Although the mecha-nisms by which baseline general characteristics influencethe efficacy of ICIs have not been fully demonstrated, theycan be used as stratification factors for efficacy and tumorprognosis in future trials to gradually expand theunderstanding.

Intestinal commensal microbiotaThe commensal microbiota plays a key role in theimmune response, with gut bacteria significantly asso-ciated with improved responses to ICIs in humans[132]. Four independent studies analyzing baselinefecal samples found that different specific intestinalbacterium were associated with response to ICIs inmelanoma [133–135], NSCLC, RCC, and urothelialcarcinoma [132]; Sivan et al. [136] reported that com-mensal bifidobacteria enhanced anti-PD-1 antibodyresponse by enhancing DC function in mice; PFS andOS after ipilimumab treatment in melanoma patientswith baseline microbiota enriched Faecalibacteriumspecies and other Firmicutes were better than thosewith baseline microbiota enriched Bacteroides [134];in addition, Routy et al. [132] revealed a correlationbetween clinical response and the relative abundanceof Akkermansia muciniphilia, which enhanced the ef-ficacy of PD-1 antibodies in an IL-12-dependent man-ner. The impact on the efficacy of ICIs may berelated to different cancer types, microbial sequencingand analysis techniques, geographical distribution ofintestinal bacteria, as well as antibiotic treatment.

Host germline geneticsHLA genes are the most polymorphic genes in the hu-man genome and encode key components of immuno-genicity. HLA-I diversity is characterized by significantsequence variation in peptide-binding regions, termedthe human immunopeptidome [137]. Analysis of 1535patients with ICI-treated tumors found that the presenceof a more diverse array of HLA-I molecules was associ-ated with increased survival [138], possibly due to itsbroader presentation of tumor antigens [139, 140]. Pa-tients who are heterozygous for all HLA-I loci in pa-tients receiving anti-PD-1 therapy have a higher on-treatment clonal expansion of TCR repertoire thanhomozygous patients [139]. A study showed that HLAloss of heterozygosity (HLA LOH) occurs in 45.1% of allpatients with advanced disease and varies by tumor type[14]. The concordance of HLA LOH detected by WESand multi-genic panels is high and suggests that HLALOH may be associated with immune escape, resultingin the resistance to immunotherapy [141]. In addition,specific HLA-I supertypes, such as the HLAB44 super-type allele, are associated with improved survival in mel-anoma patients treated with ICIs [139]. Other hostimmune-related gene polymorphisms, including HLA-IIgenes, non-classical HLA-i genes, NF-κB, and JAK-STAT family members, have also been shown to be as-sociated with tumor response to ICIs [138]. Future stud-ies are needed to further investigate the impact of hostimmune gene variation on ICIs efficacy.

(v). Immune-related adverse events

Since ICIs may cause tumor regression and immune-related adverse events (irAEs) through enhanced im-mune responses, several studies have shown a relevanceof the two. irAEs are associated with tumor regression inpatients with metastatic RCC or melanoma treated withipilimumab [142, 143]. And the early development ofoverall irAEs was associated with better ORR and PFS inNSCLC patients treated with nivolumab [144]. However,multivariate analysis by Judo [145] showed that onlylow-grade irAEs, but not high-grade irAEs, were associ-ated with better response to anti-PD-1 blockade in pa-tients with non-melanoma. In addition, different types ofirAEs are associated with immunotherapeutic responsesin different tumor types. Fujisawa et al. [146] demon-strated an association between endocrine irAEs and OSand better prognosis in melanoma patients treated withipilimumab after nivolumab; likewise, thyroid dysfunc-tion in NSCLC patients treated with anti-PD-1 was sta-tistically associated with OS and PFS [147]. Thedevelopment of vitiligo is associated with a better re-sponse to ipilimumab in melanoma patients [148], mayrepresenting a common immune response against


antigens shared by melanocytes and melanomas. Severalstudies showed that other skin irAEs were also associ-ated with better outcomes in cancer [149, 150]. But thecontroversial finding reported that the occurrence ofskin toxicities, except for vitiligo, were related to ashorter OS of melanoma patients treated with ipilimu-mab after nivolumab [146]. Since skin irAEs includevarious types of skin disorders, such as pruritus, psoria-sis, and lichenoid toxicity, the association of each skinirAE with outcome may vary. However, given that ICIsmay cause tumor regression and irAEs by enhancing im-mune responses, some biomarkers that have been ex-plored to predict the occurrence of irAEs, such as T celldiversity, cytokines and inflammatory factors, differentgut microbiome, may also be predictive biomarkers ofICIs-efficacy. Therefore, in clinical practice, how tobetter use biomarkers to achieve the best efficacy whileexperienced minimal toxicity is a difficult problem toexplore in the future.There provides an overview of predictive biomarkers

for ICIs efficacy in Fig. 1, and details of some factors inTable 1.

Exploration of predictive markers by ICI typesIn addition, considering that the type of ICIs is morecorrelated with treatment, it seems more reasonable to

Fig. 1 An overview of predictive biomarkers for immune checkpoint inhibithe efficacy of immune checkpoint inhibitors therapy are briefly describedmicroenvironment phenotype biomarkers, circulating factors, host-related f

explore biomarkers that can predict the efficacy of dif-ferent ICIs. Studies have shown an association betweentumor autoantigen expression and improved ICI-response. Eight-gene cluster known as the “anti-CTLA-4resistance associated MAGE-A (CRMA)” cluster is asso-ciated with poor response to anti-CTLA-4 rather thananti-PD-1 therapy [151]. The exact mechanism is un-known, but may be related to the idea that the expres-sion of CRMA leads to a reduction or defect inautophagy, which in turn interferes with antigen pro-cessing and presentation. Therefore, CRMA expressionis considered to be a predictive biomarker for anti-CTLA-4 therapy rather than a predictor of overall dis-ease prognosis, and CRMA gene expression may be usedto identify patients who respond to combination therapyof anti-CTLA-4 and anti-PD-1 [151]. The researchersanalyzed the expression MHC-I and II protein in tumorcells from previously untreated patients with advancedmelanoma, and correlated the results with transcrip-tomic and genomics analyses [152]. They found thatMHC proteins showed different sensitivities to CTLA-4and PD-1 blockers. Major (> 50%) or complete loss ofMHC-I expression on membranes of melanoma cell wasassociated with transcriptional repression of HLA-A,HLA-B, HLA-C, and B2M in 78/181 patients (43%),which could predict the resistance to anti-CTLA-4

tors efficacy. Key elements in predictive biomarker development forin the figure, including tumor cells-related biomarkers, tumor immuneactors, and immune-related adverse events

Table 1 Details of some factors that predict response to immune checkpoint inhibitor therapy

Type of marker Marker Association withclinical outcome

Cancer type Tissue type for markerassessment

Tumor genome and neoantigenbiomarkers

Tumor mutation burden Positive or negative Multipletumor types

Tumor tissue or blood

PD-L1 expression in tumor Positive Multipletumor types

Tumor tissue

Iindel Positive Multipletumor types

Tumor tissue

SCNAs Positive Multipletumor types

Tumor tissue

DNAm, e.g., HOX gene methylation Positive NSCLC Tumor tissue or blood

DDR pathways, e.g., dMMR/MSI, BER,HRR;

Positive Multipletumor types

Tumor tissue

IFN-γ pathway genes, IFNGR1/2, JAK1/2,and IRF1

Negative Multipletumor types


STK11 Positive or Unknown NSCLC Tumor tissue or blood

Neoantigen load, low neoantigenintratumour heterogeneity

Positive Multipletumor type

Tumor tissue

Tumor immune microenvironmentphenotype biomarkers

PD-L1 expression in TME Positive Multipletumor types

Tumor tissue

Immune-inflamed TME Positive Multipletumor types

Tumor tissue

T cell repertoire clonality Positive Multipletumor types


CD39 + CD8 + TIL Positive NSCLC, RCC Tumor tissue or blood

Liquid biopsy biomarkers NLR Negative Melanoma,NSCLC

Blood

High mutation number of ctDNA orfavorable ctDNA profiles

Positive Multipletumor types

Blood

LDH Negative Melanoma Blood

IL-8 Negative Multipletumor types

Blood

Exosomal PD-L1 Positive or negative Melanoma,NSCLC

Blood

PD-L1 variant fragments Negative NSCLC Blood

Host-related markers Gender Male: positive Multipletumor types

–

Age Positive or negative orUnknown

Multipletumor types

–

Body fat distribution Positive Multipletumor types

–

Specific Intestinal microbiota Positive or negative Multipletumor types

Oral or gut

HLA-I diversity Positive Melanoma,NSCLC

Blood

HLA LOH Negative Melanoma Tumor tissue

irAEs irAEs in different organs Positive or Unknown Multipletumor types

–

PD-L1 programmed cell death-ligand 1, RCC renal cell carcinoma, NSCLC non-small cell lung cancer, TMB tumor mutation burden, indel insertion and deletion,SCNAs somatic copy number alterations, MMR mismatch repair, dMMR MMR deficiency, MSI microsatellite instability, TIL tumor infiltrating lymphocyte, POLE/POLD1polymerase gene epsilon/delta 1, BER base excision repair, HRR homologous recombination repair, DDR DNA damage response, HLA human leukocyte antigen,TME tumor microenvironment, NLR neutrophil-to-lymphocyte ratio, ctDNA circulating tumor DNA, IL-8 interleukin-8, LDH lactate dehydrogenase, irAE immune-related adverse event, DNAm DNA methylation HLA LOH HLA loss of heterozygosity



antibody therapy but not anti-PD-1 therapy. MHC-II ex-pression was observed in > 1% of melanoma cells in 55/181 (30%) patients, and correlated with IFN-γ and itsmediated gene signature, which could predict the re-sponse to anti-PD-1 but not anti-CTLA-4 therapy [152].Thus, MHC-I expression is required for the primary re-sponse against CTLA-4 for melanoma, while the primaryresponse to anti-PD-1 is associated with pre-existingIFN-γ-mediated immune activation. Therefore, the ex-ploration of markers to predict the therapeutic efficacyor resistance of different ICIs is also essential. Morestudies are expected in the future to analyze the mecha-nisms of action of different ICIs and their interactionswith tumors in depth.

Comprehensive predictors of ICIs efficacyThe current understanding of the clinical response toICIs-treatment suggests that any single biomarker can-not effectively identify the benefit populations. The spe-cificity and efficacy of prediction will be greatlyimproved when combination of multiple factors is usedas a composite variable to capture immune status. Rizviet al. [153] found that TMB and PD-L1 were two inde-pendent factors affecting the efficacy of immunotherapy,while patients with both high levels of TMB and positivePD-L1 had the highest duration of benefit rate; anotherstudy showed that NSCLC patients with both high TILdensity and high PD-L1 expression treated with PD-L1inhibitor had the highest positive predictive value ofORR and the longest PFS [154]; and Yu et al. [155] fur-ther demonstrated that the comprehensive variables ofthree predictive markers, CD8+TIL, PD-L1 expression,and TMB, were associated with improved OS and PFScompared with a single biomarker or two of the threebiomarkers. Furthermore, the use of big data analysis topredict markers of immunotherapy efficacy helps to es-tablish a new framework for precise treatment of tu-mors. A study of 4 groups of clinical trials covering 22cancer types and more than 300 patients evaluated therelationship between biomarkers and best overallresponses (BOR), PFS. It was found that TMB, T cell-inflamed GEPs were associated with the efficacy of clin-ical immunotherapy, and the higher TMB, the higherORR [156].In addition, developing predictive models by integrat-

ing different types of data based on different compo-nents of tumor-host interactions seems to have a goodprospect. A research team created two neoantigen im-mune fitness models by computational biology methods,namely, the neoantigen quantity model, mainly statisti-cally analyzing the number of tumor antigens, and theneoantigen quality fitness model, involving various fac-tors such as the similarity between tumor antigens andpathogen antigens and the binding ability to TCR [157].

The results showed that only the neoantigen quality fit-ness model could better predict the postoperative sur-vival of patients with pancreatic cancer. Another studydeveloped a new neoantigen fitness model includingthree elements (tumor clonality, DAI, and microbial epi-tope homology), which was quantified as a nonlinearfunction of alignment scores, and the results showedthat the model incorporating all three elements success-fully predicted survival in all three ICI-treatment cohorts[60]. But before applying the model more broadly, it isnecessary to identify unique parameters for each cancerspecies and/or therapeutic agent [158]. Jiang et al. [159]designed a completely new computational architecture,TIDE score ratio biomarkers (tumor mutation load, PD-L1 level, and INF-γ), namely tumor immune dysfunctionand rejection scores. It reveals the impact of tumor infil-tration levels of different immune cell types on overallsurvival of patient by analyzing the TCGA and PRECOGdatabases and synthesizing different types of tumor im-mune escape mechanisms. Using this framework andpretreatment RNA-Seq or NanoString tumor expressionprofiling, they have identified that TIDE more accuratelypredicts the outcome of melanoma patients treated withfirst-line anti-PD-1 or anti-CTLA-4 than other bio-markers such as PD-L1 levels and TMB. TIDE also re-vealed novel candidate regulators of resistance to ICIs,such as SERPINB9, demonstrating utility for immuno-therapeutic studies.The combined detection of independent predictive

markers makes more patients to receive ICIs and ex-pands the beneficiary population, while for interactingmarkers, a bioinformatics-based predictive-model can beestablished according to different impact weights of eachfactor and improve the accuracy of screening the benefi-ciary population by comprehensive consideration, andhow to better utilize the interrelationship network ofvarious markers is an aspect to be considered of com-prehensive predictive-models; in addition, it should beexplored how the combined prediction with multiplefactors achieve the optimal cost-effectiveness to servethe clinical immunotherapy of tumors more effectively.In the future, it may be promising to obtain the most ef-fectively comprehensive predictive-markers by extractingfeatures with large samples and multiple dimensions andconstructing multivariate models using machine learningand artificial intelligence.

Summary and outlookIn this review, we deeply analyze the exploration courseand research progress of predictive biomarkers as an ad-junctive tool to tumor immunotherapy in identifyingICIs efficacy. In recent years, predictive markers of ICIsefficacy have been gradually explored from the expres-sion of intermolecular interactions within tumor cells to


the expression of various molecules and cells in TME,even been extended to the exploration of circulating andhost systemic markers, and gradually realized theprocess from the identification of single marker to thedevelopment of multifactorial synergistic predictivemarkers. The exploration of predictive biomarkers ofICIs efficacy indicates a complex interaction between theregulation of the immune system network and tumors,reflecting more comprehensively the complexity and di-versity of the effects of immunotherapy on tumors andeven the whole body. Nonetheless, the findings of somebiomarkers explored in the review are contradictory andthe mechanisms of action are not well understood,which need to be confirmed by further large-scale pro-spective studies, but these breakthrough findings offer agreat promise for biomarker strategies with more accur-ate positive and negative predictive values that can beused routinely in clinical practice to assist patients withdifferent malignancies in ICIs-based therapy manage-ment, monitor disease development, and conquer tumorresistant to immunotherapy.With the development of basic technology research

such as multiplex IHC, high-throughput sequencingtechnology, and microarray technology, more and morepotential markers can be widely screened widely on agenomic scale and a variety of proteins and cell popula-tions can be quantified. However, several knowledgegaps still exist. First, for single marker, further cognitionof PD-L1 and TMB should be enhanced, while continu-ing to promote consistency evaluation of detectionmethods; Gene mutations show great potential in theevaluation and monitoring of the whole course of im-munotherapy and should be continuously explored;There are still many unknowns about the exploration ofmarkers of the immune microenvironment and hostmicroenvironment, which needs to be understood froma deeper molecular perspective. Secondly, in view of thevarious emerging biomarkers and the disadvantages ofevery single marker to varying degrees, strategies com-bining two or more approaches to capture immune sta-tus may be more effective as composite predictivebiomarkers for ICIs efficacy. The advantages of eachmarker should be fully utilized to lay the foundation forthe development of multifactorial predictive models. Bal-ancing the relationship between the scientificity, accessi-bility, and simple operation of the clinical application ofeach predictive marker/model is a challenge to considerin clinical research. Thirdly, the exploration of moresimple and feasible prediction means in clinical practice.For example, the potential of liquid biopsy such asctDNA in the whole process of efficacy evaluation andmonitoring of immunotherapy should be fully devel-oped. Predicting the long-term survival of immunother-apy based on biomarkers in peripheral blood is a

potential development direction. Furthermore, the use ofmachine deep learning and artificial intelligence to ex-plore the mechanisms and markers of immunotherapyefficacy and drug resistance is changing from fantasy toreality, which can be used as the direction of future sci-entific research and clinical exploration. Multivariatepredictive models need to extract data features with largesamples and multiple dimensions using machine learning,and integrate different types of data based on differentcomponents of tumor-host interactions for comprehen-sive validation and evaluation, including polymorphismdata such as intratumoral genomic and molecular charac-teristics, tumor immune microenvironment phenotype,peripheral blood biomarkers and host-related factors. Fi-nally, given that multiple patterns of atypical response,such as pseudoprogression, occur during immunotherapy,which significantly affect patient treatment and overallsurvival, it is also essential for the exploration of predictivemarkers to these special response pattern. In the future,through the scientific study of the availability of multiplemarkers and the exploration of feasibility and reproduci-bility in clinical practice, standardized predictive bio-markers (models) for ICIs response would be establishedto maximize the benefit of patients from these transforma-tive treatments, ultimately prompting the field to developtowards precision immuno-oncology.

AbbreviationsCAR: chimeric antigen receptor; BOR: best overall responses; ICI: Immunecheckpoint inhibitor; PD-1/PD-L1: Programmed cell death-1/programmedcell death-ligand 1; RCC: Renal cell carcinoma; HNSCC: Head and necksquamous cell carcinoma; ORR: Objective response rate; NSCLC: Non-smallcell lung cancer; TMB: Tumor mutation burden; SCLC: Small cell lung cancer;HPV: Human papilloma virus; NCCN: National Comprehensive CancerNetwork; OS: Overall survival; DCR: Disease control rate; DCB: Durable clinicalbenefit; PFS: Progression-free survival; CGP: Comprehensive genomicprofiling; NGS: Next generation sequencing; AF: Allele frequency; CNV: Copynumber variation; nsSNV: Nonsynonymous single nucleotide variant;SCNAs: Somatic copy number alterations; indel: Insertion and deletion;TCGA: The Cancer Genome Atlas; MMR: Mismatch repair; dMMR: MMRdeficiency; MSI: Microsatellite instability; TIL: Tumor infiltrating lymphocyte;IDO: Indoleamine 2,3-dioxygenase; POLE/POLD1: Polymerase gene epsilon/delta 1; BER: Base excision repair; HRR: Homologous recombination repair;DDR: DNA damage response; dsRNA: Double-stranded RNA;ADAR1: Adenosine deaminase acting on RNA; MDA5: Melanomadifferentiation-associated protein 5; CTLA-4: Cytotoxic T-lymphocyte-associated protein-4; EGFR: Epidermal growth factor receptor; ALK: Anaplasticlymphoma kinase; FDA: Food and Drug Administration; AACR: AmericanAssociation for Cancer Research; ERV: Endogenous retroviruse; MHC: Majorhistocompatibility complex; DAI: Differential agretopicity index;IEDB: Immune Epitope Database; HLA: Human leukocyte antigen;IHC: Immunohistochemistry; TPS: Tumor proportion score; CPS: Combinedpositive score; IPS: Immune positive score; TIM-3: T cell immunoglobulin-3;LAG-3: Lymphocyte activation gene-3; VISTA: V-domain Ig suppressor of T-cell activation; TME: Tumor microenvironment; ROC curve: Receiver operatingcharacteristic curve; scRNA-seq: Single-cell mRNA sequencing;TCF7: Transcription factor 7; FcγR: Fc domain glycan of the drug and Fcγreceptor; TAM: Tumor-associated macrophage; NK cell: Natural killer cell;TCR: T cell receptor; IR: Immune repertoire; GEP: Gene expression profile;NLR: Neutrophil-to-lymphocyte ratio; ICOS: Inducible T Cell Co-stimulator;CTC: Circulating tumor cell; ctDNA: Circulating tumor DNA; MDSC: Myeloid-derived suppressor cell; TNF: Tumor necrosis factor; IL: Interleukin; CRP: C-reactive protein; LDH: Lactate dehydrogenase; RCT: Randomized controlled


trial; PS: Performance status; ASCO: American Society of Clinical Oncology;DOR: Duration of response; BMI: Body mass index; irAE: immune-relatedadverse event; WES: whole exome sequencing; HLA LOH: HLA loss ofheterozygosity; dNLR: derived NLR; LIPI: lung immune prognostic index;FGFR2: fibroblast growth factor receptor 2; METex14: MET exon 14 skipping;CRMA: CTLA-4 resistance associated MAGE-A

AcknowledgementsNot applicable.

Authors’ contributionsRL Bai carried out the primary literature search, drafted and revised themanuscript, and participated in discussions. Z Lv and DS Xu helped modifythe manuscript. JW Cui carried out the design of the research and literatureanalysis, drafted and revised the manuscript, and participated in discussions.All authors read and approved the final manuscript.

FundingThis work was supported by the National Natural Science Foundation ofChina (Grant 81672275 and 81874052); Project of Jilin Provincial Departmentof Education (Grant JJKH20190020KJ); Project of Department of Science andTechnology of Jilin Province (Grants 20180101009JC and 20190303146SF);

Availability of data and materialsNot applicable.

Ethics approval and consent to participateNot applicable.

Consent for publicationNot applicable.

Competing interestsThe authors declare that there are no conflicts of interest.

Received: 4 June 2020 Accepted: 29 July 2020

References1. Gong J, Chehrazi-Raffle A, Reddi S, et al. Development of PD-1 and PD-L1

inhibitors as a form of cancer immunotherapy: a comprehensive review ofregistration trials and future considerations [J]. Immunother Cancer. 2018;6(1):8.

2. Akinleye A, Rasool Z. Immune checkpoint inhibitors of PD-L1 as cancertherapeutics [J]. J Hematol Oncol. 2019;12(1):92.

3. Robert C, Long GV, Brady B, et al. Nivolumab in previously untreatedmelanoma without BRAF mutation [J]. N Engl J Med. 2015;372(4):320–30.

4. Garon EB, Rizvi NA, Hui R, et al. Pembrolizumab for the treatment of non-small-cell lung cancer [J]. N Engl J Med. 2015;372(21):2018–28.

5. Brahmer J, Reckamp KL, Baas P, et al. Nivolumab versus Docetaxel inadvanced squamous-cell non-small-cell lung Cancer [J]. N Engl J Med. 2015;373(2):123–35.

6. Legrand FA, Gandara DR, Mariathasan S, et al. Association of high tissueTMB and atezolizumab efficacy across multiple tumor types [J]. J Clin Oncol.2018;36(15_suppl):12000.

7. Rosenberg JE, Hoffman-Censits J, Powles T, et al. Atezolizumab in patientswith locally advanced and metastatic urothelial carcinoma who haveprogressed following treatment with platinum-based chemotherapy: asingle-arm, multicentre, phase 2 trial [J]. Lancet. 2016;387(10031):1909–20.

8. Hellmann MD, Callahan MK, Awad MM, et al. Tumor mutational burden andefficacy of Nivolumab Monotherapy and in combination with Ipilimumab insmall-cell lung Cancer [J]. Cancer Cell. 2018;33(5):853–61.e4.

9. Hellmann MD, Nathanson T, Rizvi H, et al. Genomic features of response tocombination immunotherapy in patients with advanced non-small-cell lungCancer [J]. Cancer Cell. 2018;33(5):843–52.e4.

10. Hellmann MD, Ciuleanu TE, Pluzanski A, et al. Nivolumab plus Ipilimumab inLung Cancer with a High Tumor Mutational Burden [J]. N Engl J Med. 2018;378(22):2093–104.

11. Singal G, Miller PG, Agarwala V, et al. Association of Patient Characteristicsand Tumor Genomics with Clinical Outcomes among Patients with non-

Small Cell Lung Cancer Using a Clinicogenomic database [J]. Jama. 2019;321(14):1391–9.

12. Goodman AM, Kato S, Bazhenova L, et al. Tumor mutational burden as anindependent predictor of response to immunotherapy in diverse cancers [J].Mol Cancer Ther. 2017;16(11):2598–608.

13. Hanna GJ, Lizotte P, Cavanaugh M, et al. Frameshift events predict anti-PD-1/L1 response in head and neck cancer [J]. JCI Insight. 2018;3(4):e98811.

14. Yarchoan M, Hopkins A, Jaffee EM. Tumor mutational burden and responserate to PD-1 inhibition [J]. N Engl J Med. 2017;377(25):2500–1.

15. Miao D. Margolis CA. Genomic correlates of response to immunecheckpoint therapies in clear cell renal cell carcinoma [J]. Science. 2018;359(6377):801–6.

16. Riaz N, Havel JJ, Makarov V, et al. Tumor and Microenvironment Evolutionduring Immunotherapy with Nivolumab [J]. Cell. 2017;171(4):934–49.e16.

17. Fuchs C, ÖZgÜRoĞLu M, Bang Y-J, et al. The association of molecularbiomarkers with efficacy of pembrolizumab versus paclitaxel in patientswith gastric cancer (GC) from KEYNOTE-061[J]. J Clin Oncol. 2020;38:4512.

18. Shitara K, ÖZgÜRoĞLu M, Bang Y-J, et al. The association of tissue tumormutational burden (tTMB) using the Foundation Medicine genomicplatform with efficacy of pembrolizumab versus paclitaxel in patients (pts)with gastric cancer (GC) from KEYNOTE-061[J]. J Clin Oncol. 2020;38:4537.

19. Li W, Matakidou A, Ghazoui Z, et al. Molecular biomarkers to identifypatients (pts) who may benefit from durvalumab (D; anti-PD-L1) ±tremelimumab (T;anti-CTLA-4) in recurrent/metastatic head and necksquamous cell carcinoma (R/M HNSCC) from HAWK and CONDOR studies[J].J Clin Oncol. 2020;38:6548.

20. Li W, Wildsmith S, Ye J, et al. Plasma-based tumor mutational burden(bTMB) as predictor for survival in phase III EAGLE study: Durvalumab (D) ±tremelimumab (T) versus chemotherapy (CT) in recurrent/metastatic headand neck squamous cell carcinoma (R/M HNSCC) after platinum failure[J]. JClin Oncol. 2020;38:6511.

21. Kawazoe A, Yamamoto N, Kotani D, et al. TAS-116, an oral HSP90 inhibitor,in combination with nivolumab in patients with colorectal cancer and othersolid tumors: An open-label, dose-finding, and expansion phase Ib trial(EPOC1704)[J]. J Clin Oncol. 2020;38:4044.

22. Kowanetz M, Zou W, Shames D, et al. OA20.01 tumor mutation burden(TMB) is associated with improved efficacy of Atezolizumab in 1L and 2L+NSCLC patients [J]. J Thorac Oncol. 2017;12(1):S321–2.

23. Rosenberg JE, et al. Atezolizumab as first-line treatment in cisplatin-ineligible patients with locally advanced and metastatic urothelialcarcinoma: a singlearm, multicentre, phase 2 trial. Lancet. 2017;389:67.

24. Samstein RM, Lee CH, Shoushtari AN. Tumor mutational load predictssurvival after immunotherapy across multiple cancer types[J]. Nat Genet.2019;51(2):202–6.

25. Johnson DB, Frampton GM, Rioth MJ, et al. Targeted next generationsequencing identifies markers of response to PD-1 blockade [J]. CancerImmunol Res. 2016;4(11):959–67.

26. Ramalingam S, Hellmann M, Awad M, et al. Abstract CT078: Tumormutational burden (TMB) as a biomarker for clinical benefit from dualimmune checkpoint blockade with nivolumab (nivo) + ipilimumab (ipi) infirst-line (1L) non-small cell lung cancer (NSCLC): identification of TMB cutofffrom CheckMate 568[J]. Cancer Res. 2018;78:CT078.

27. Fang W, Ma Y, Yin JC, et al. Comprehensive genomic profiling identifiesnovel genetic predictors of response to anti-PD-(L)1 therapies in Non-SmallCell Lung Cancer[J].Clin Cancer Res. 2019;25(16):5015–26.

28. Wang Z, Duan J, Wang G, et al. Allele frequency-adjusted blood-basedtumor mutational burden as a predictor of overall survival for patients withNSCLC treated with PD-(L)1 inhibitors [J]. J Thorac Oncol. 2020;15(4):556–67.

29. Cai L, Bai H, Duan J, et al. Epigenetic alterations are associated with tumormutation burden in non-small cell lung cancer [J]. J Immunother Cancer.2019;7(1):198.

30. Turajlic S, Litchfield K, Xu H, et al. Insertion-and-deletion-derived tumour-specific neoantigens and the immunogenic phenotype: A pan-canceranalysis [J]. Lancet Oncol. 2017;18(8):1009–21.

31. Davoli T, Uno H, Wooten EC, et al. Tumor aneuploidy correlates withmarkers of immune evasion and with reduced response to immunotherapy[J]. Science. 2017;355(6322):eaaf8399.

32. Taylor AM, Shih J, Ha G, et al. Genomic and Functional Approaches toUnderstanding Cancer Aneuploidy [J]. Cancer Cell. 2018;33(4):676–89.e3.

33. Zhang Y, Sun Z, Mao X, et al. Impact of mismatch-repair deficiency on the colorectalcancer immune microenvironment [J]. Oncotarget. 2017;8(49):85526–36.


34. Overman MJ, Lonardi S, Wong KYM, et al. Durable clinical benefit withNivolumab plus Ipilimumab in DNA mismatch repair-deficient/microsatellite instability-high metastatic colorectal Cancer [J]. J ClinOncol. 2018;36(8):773–9.

35. Wang F, Zhao Q, Wang YN, et al. Evaluation of POLE and POLD1 mutationsas biomarkers for immunotherapy outcomes across multiple Cancer types[J]. JAMA Oncol. 2019;5(10):1504–6.

36. Wang Z, Zhao J, Wang G, et al. Comutations in DNA damage responsepathways serve as potential biomarkers for immune checkpoint blockade[J]. Cancer Res. 2018;78(22):6486–96.

37. Possick JD. Pulmonary toxicities from checkpoint immunotherapy formalignancy [J]. Clin Chest Med. 2017;38(2):223–32.

38. Darnell J, Kerr I, Stark G. Jak-STAT pathways and transcriptional activation inresponse to IFNs and other extracellular signaling proteins [J]. Science. 1994;264(5164):1415–21.

39. Wang X, Haswell JR, Roberts CWM. Molecular pathways: SWI/SNF (BAF)complexes are frequently mutated in cancer--mechanisms and potentialtherapeutic insights [J]. Clin Cancer Res. 2014;20(1):21–7.

40. Miao D, Margolis CA, Gao W, et al. Genomic correlates of response toimmune checkpoint therapies in clear cell renal cell carcinoma [J]. Science.2018;359(6377):801–6.

41. Pan D, Kobayashi A, Jiang P, et al. A major chromatin regulator determinesresistance of tumor cells to T cell-mediated killing [J]. Science (New York, N.Y.). 2018;359(6377):770–5.

42. Ishizuka JJ, Manguso RT, Cheruiyot CK, et al. Loss of ADAR1 in tumoursovercomes resistance to immune checkpoint blockade [J]. Nature. 2019;565(7737):43–8.

43. De Velasco G, Je Y, Bosse D, et al. Comprehensive meta-analysis of keyimmune-related adverse events from CTLA-4 and PD-1/PD-L1 inhibitors inCancer patients [J]. Cancer Immunol Res. 2017;5(4):312–8.

44. Chae YK, Davis AA, Raparia K, et al. Association of Tumor Mutational BurdenWith DNA Repair Mutations and Response to Anti-PD-1/PD-L1 Therapy inNon-Small-Cell Lung Cancer [J]. Clin Lung Cancer. 2019;20(2):88-96.e6.

45. Rizvi H, Sanchez-Vega F, La K, et al. Molecular determinants of response toanti-programmed cell death (PD)-1 and anti-programmed death-ligand(PD-L)-ligand 1 blockade in patients with non-small-cell lung Cancerprofiled with targeted next-generation sequencing [J]. J Clin Oncol.2018;36(7):633–41.

46. Klempner SJ, Fabrizio D, Bane S. Tumor Mutational Burden as a PredictiveBiomarker for Response to Immune Checkpoint Inhibitors: A Review ofCurrent Evidence [J]. Oncologist. 2020;25(1):e147–59.

47. Skoulidis F, Goldberg ME. STK11/LKB1 Mutations and PD-1 InhibitorResistance in KRAS-Mutant Lung Adenocarcinoma [J]. Cancer Discov. 2018;8(7):822–35.

48. Ross JS, Goldberg ME, Albacker LA, et al. 1138PDImmune checkpointinhibitor (ICPI) efficacy and resistance detected by comprehensive genomicprofiling (CGP) in non-small cell lung cancer (NSCLC)[J]. Annals of Oncology.2017;28:mdx376.004.

49. Zhang K, Hong X, Song Z, et al. Identification of deleterious NOTCHmutation as novel predictor to efficacious immunotherapy in NSCLC [J].2020;26(14):3649–61.

50. Rooney MS, Shukla SA, Wu CJ, et al. Molecular and genetic properties oftumors associated with local immune cytolytic activity[J]. Cell. 2015;160(1-2):48–61.

51. Smith CC, Beckermann KE, Bortone DS, et al. Endogenous retroviralsignatures predict immunotherapy response in clear cell renal cellcarcinoma [J]. J Clin Invest. 2018;128(11):4804–20.

52. Tamiya Y, Zenke Y, Matsumoto S, et al. Abstract 9589: Therapeutic impact ofmutation subtypes and concomitant STK11 mutations in KRAS –mutatednonsmall cell lung cancer (NSCLC): A result of nationwide genomicscreening project (LC-SCRUM-Japan)[J]. J Clin Oncol. 2020;38(15_suppl):9589.

53. McGranahan N, Furness AJ, Rosenthal R, et al. Clonal neoantigens elicit Tcell immunoreactivity and sensitivity to immune checkpoint blockade [J].Science. 2016;351(6280):1463–9.

54. Jiang T, Shi T, Zhang H, et al. Tumor neoantigens: from basic research toclinical applications [J]. J Hematol Oncol. 2019;12(1):93.

55. Yi M, Qin S, Zhao W, et al. The role of neoantigen in immune checkpointblockade therapy [J]. Exp Hematol Oncol. 2018;7:28.

56. Rizvi NA, Hellmann MD, Snyder A, et al. Cancer immunology. Mutationallandscape determines sensitivity to PD-1 blockade in non-small cell lungcancer [J]. Science. 2015;348(6230):124–8.

57. Duan F, Duitama J, Al Seesi S, et al. Genomic and bioinformatic profiling ofmutational neoepitopes reveals new rules to predict anticancerimmunogenicity [J]. J Exp Med. 2014;211(11):2231–48.

58. Ghorani E, Rosenthal R, McGranahan N, et al. Differential binding affinity ofmutated peptides for MHC class I is a predictor of survival in advanced lungcancer and melanoma [J]. Ann Oncol. 2018;29(1):271–9.

59. Rech AJ, Balli D, Mantero A, et al. Tumor immunity and survival as afunction of alternative Neopeptides in human Cancer [J]. Cancer ImmunolRes. 2018;6(3):276–87.

60. Łuksza M, Riaz N, Makarov V, et al. A neoantigen fitness model predictstumour response to checkpoint blockade immunotherapy [J]. Nature. 2017;551(7681):517–20.

61. Kim S, Kim HS, Kim E, et al. Neopepsee: accurate genome-level prediction ofneoantigens by harnessing sequence and amino acid immunogenicityinformation [J]. Ann Oncol. 2018;29(4):1030–6.

62. Yi M, Dong B, Chu Q, et al. Immune pressures drive the promoterhypermethylation of neoantigen genes [J]. Exp Hematol Oncol. 2019;8:32.

63. Wolchok JD, Chiarionsileni V, Gonzalez R, et al. Updated results from aphase III trial of nivolumab (NIVO) combined with ipilimumab (IPI) intreatmentnaive patients (pts) with advanced melanoma (MEL) (CheckMate067)[J]. J Clin Oncol. 2016;34:9505.

64. Gandhi L, Rodríguez-Abreu D, Gadgeel S, et al. Pembrolizumab plusChemotherapy in Metastatic Non–Small-Cell Lung Cancer [J]. N Engl J Med.2018;378(22):2078–92.

65. Rouquette I, Taranchon-Clermont E, Gilhodes J, et al. Immune biomarkers inthymic epithelial tumors: expression patterns, prognostic value andcomparison of diagnostic tests for PD-L1[J]. Biomark Res. 2019;7:28.

66. Topalian SL, Taube JM, Anders RA, et al. Mechanism-driven biomarkers toguide immune checkpoint blockade in cancer therapy [J]. Nat Rev Cancer.2016;16(5):275–87.

67. Gibney GT, Weiner LM, Atkins MB. Predictive biomarkers for checkpointinhibitor-based immunotherapy [J]. Lancet Oncol. 2016;17(12):e542–51.

68. Carbone DP, Reck M, Paz-Ares L, et al. First-line Nivolumab in stage IV orrecurrent non-small-cell lung Cancer [J]. N Engl J Med. 2017;376(25):2415–26.

69. Wu K, Yi M, Qin S, et al. The efficacy and safety of combination of PD-1 andCTLA-4 inhibitors: a meta-analysis [J]. Exp Hematol Oncol. 2019;8:26.

70. Sacher AG, Gandhi L. Biomarkers for the clinical use of PD-1/PD-L1 inhibitorsin non-small-cell lung Cancer: a review [J]. JAMA Oncol. 2016;2(9):1217–22.

71. Nishino M, Ramaiya NH, Hatabu H, et al. Monitoring immune-checkpointblockade: response evaluation and biomarker development [J]. Nat Rev ClinOncol. 2017;14(11):655–68.

72. Hong L, Negrao MV, Dibaj SS, et al. Programmed Death Ligand 1Heterogeneity and its Impact on Benefit from Immune CheckpointInhibitors in Non-Small-Cell Lung Cancer.[J]. Ann Oncol. 2019;30(8):1311–20.

73. Herbst RS, Soria JC, Kowanetz M, et al. Predictive correlates of response tothe anti-PD-L1 antibody MPDL3280A in cancer patients [J]. Nature. 2014;515(7528):563–7.

74. Chen DS, Mellman I. Elements of cancer immunity and the cancer-immuneset point [J]. Nature. 2017;541(7637):321–30.

75. Llosa NJ, Cruise M, Tam A, et al. The vigorous immune microenvironment ofmicrosatellite instable colon cancer is balanced by multiple counter-inhibitory checkpoints [J]. Cancer Discov. 2015;5(1):43–51.

76. Joyce JA, Fearon DT. T cell exclusion, immune privilege, and the tumormicroenvironment [J]. Science. 2015;348(6230):74–80.

77. Tumeh PC, Harview CL, Yearley JH, et al. PD-1 blockade induces responsesby inhibiting adaptive immune resistance [J]. Nature. 2014;515(7528):568–71.

78. Galon J, Fox BA, Bifulco CB, et al. Immunoscore and Immunoprofiling incancer: an update from the melanoma and immunotherapy bridge 2015[J].J Transl Med. 2016;14:273.

79. Bindea G, Mlecnik B, Angell HK, et al. The immune landscape of humantumors: implications for cancer immunotherapy [J]. Oncoimmunology. 2014;3(1):e27456.

80. Mlecnik B, Tosolini M, Kirilovsky A, et al. Histopathologic-based prognosticfactors of colorectal cancers are associated with the state of the localimmune reaction [J]. J Clin Oncol. 2011;29(6):610–8.

81. Ji RR, Chasalow SD, Wang L, et al. An immune-active tumormicroenvironment favors clinical response to ipilimumab [J]. CancerImmunol Immunother. 2012;61(7):1019–31.

82. Ascierto ML, Kmieciak M, Idowu MO, et al. A signature of immune functiongenes associated with recurrence-free survival in breast cancer patients [J].Breast Cancer Res Treat. 2012;131(3):871–80.


83. Ribas A, Robert C, Hodi FS, et al. Association of response to programmeddeath receptor 1 (PD-1) blockade with pembrolizumab (MK-3475) with aninterferon-inflammatory immune gene signature [abstract][J]. Polym Int.2015;33(2):141–7.

84. Sade-Feldman M, Yizhak K, Bjorgaard SL, et al. Defining T Cell StatesAssociated with Response to Checkpoint Immunotherapy in Melanoma [J].Cell. 2018;175(4):998–1013 e1020.

85. Balatoni T, Mohos A, Papp E, et al. Tumor-infiltrating immune cells aspotential biomarkers predicting response to treatment and survival inpatients with metastatic melanoma receiving ipilimumab therapy [J]. CancerImmunol Immunother. 2018;67(1):141–51.

86. Simoni Y, Becht E, Fehlings M, et al. Bystander CD8(+) T cells are abundantand phenotypically distinct in human tumour infiltrates [J]. Nature. 2018;557(7706):575–9.

87. Sean, Arlauckas, Christopher, et al. In vivo imaging reveals a tumor-associated macrophage-mediated resistance pathway in anti-PD-1 therapy[J]. Sci Transl Med. 2017;9(389):eaal3604.

88. Neubert NJ, Schmittnaegel M. T cell-induced CSF1 promotes melanomaresistance to PD1 blockade [J]. Sci Transl Med. 2018;10(436):eaan3311.

89. McDermott DF, Huseni MA, Atkins MB, et al. Clinical activity and molecularcorrelates of response to atezolizumab alone or in combination withbevacizumab versus sunitinib in renal cell carcinoma [J]. Nat Med. 2018;24(6):749–57.

90. Roh W, Chen PL. Integrated molecular analysis of tumor biopsies onsequential CTLA-4 and PD-1 blockade reveals markers of response andresistance [J]. Sci Transl Med. 2017;9(379):eaah3560.

91. Forde PM, Chaft JE, Smith KN, et al. Neoadjuvant PD-1 blockade inResectable. Lung Cancer [J]. N Engl J Med. 2018;378(21):1976–86.

92. Inoue H, Park JH, Kiyotani K, et al. Intratumoral expression levels of PD-L1,GZMA, and HLA-A along with oligoclonal T cell expansion associate withresponse to nivolumab in metastatic melanoma [J]. Oncoimmunology.2016;5(9):e1204507.

93. Hopkins AC, Yarchoan M, Durham JN, et al. T cell receptor repertoirefeatures associated with survival in immunotherapy-treated pancreaticductal adenocarcinoma [J]. JCI Insight. 2018;3(13):e122092.

94. Han J, Duan J, Bai H, et al. TCR repertoire diversity of peripheral PD-1(+)CD8(+) T cells predicts clinical outcomes after immunotherapy in patientswith non-small cell lung Cancer [J]. Cancer Immunol Res. 2020;8(1):146–54.

95. Delyon J, Mateus C, Lefeuvre D, et al. Experience in daily practice withipilimumab for the treatment of patients with metastatic melanoma: anearly increase in lymphocyte and eosinophil counts is associated withimproved survival [J]. Ann Oncol. 2013;24(6):1697–703.

96. Bagley SJ, Kothari S, Aggarwal C, et al. Pretreatment neutrophil-to-lymphocyte ratio as a marker of outcomes in nivolumab-treated patientswith advanced non-small-cell lung cancer [J]. Lung Cancer. 2017;106:1–7.

97. Weide B, Martens A, Hassel JC, et al. Baseline biomarkers for outcome ofmelanoma patients treated with Pembrolizumab [J]. Clin Cancer Res. 2016;22(22):5487–96.

98. Martens A, Wistuba-Hamprecht K, Geukes Foppen M, et al. Baselineperipheral blood biomarkers associated with clinical outcome of advancedmelanoma patients treated with Ipilimumab [J]. Clin Cancer Res. 2016;22(12):2908–18.

99. Kazandjian D, Gong Y, Keegan P, et al. Prognostic value of the lung immuneprognostic index for patients treated for metastatic non-small cell lungCancer [J]. JAMA Oncol. 2019;5(10):1481–5.

100. Tietze JK, Angelova D, Heppt MV, et al. The proportion of circulatingCD45RO(+)CD8(+) memory T cells is correlated with clinical response inmelanoma patients treated with ipilimumab [J]. Eur J Cancer. 2017;75:268–79.

101. Subrahmanyam PB, Dong Z, Gusenleitner D, et al. Distinct predictivebiomarker candidates for response to anti-CTLA-4 and anti-PD-1immunotherapy in melanoma patients [J]. J Immunother Cancer. 2018;6(1):18.

102. Krieg C, Nowicka M, Guglietta S, et al. High-dimensional single-cell analysispredicts response to anti-PD-1 immunotherapy [J]. Nat Med. 2018;24(2):144–53.

103. Liakou CI, Kamat A, Tang DN, et al. CTLA-4 blockade increases IFNgamma-producing CD4+ICOShi cells to shift the ratio of effector to regulatory Tcells in cancer patients [J]. Proc Natl Acad Sci U S A. 2008;105(39):14987–92.

104. Kulasinghe A, Perry C, Kenny L, et al. D-L1 expressing circulating tumourcells in head and neck cancers [J]. BMC Cancer. 2017;17(1):333.

105. Heitzer E, Ulz P, Geigl JB. Circulating tumor DNA as a liquid biopsy forCancer [J]. Clin Chem. 2015;61(1):112–23.

106. Khagi Y, Goodman AM, Daniels GA, et al. Hypermutated Circulating TumorDNA: Correlation with Response to Checkpoint Inhibitor-BasedImmunotherapy[J]. Clin Cancer Res. 2017;23(19):5729–36.

107. Lee JH, Long GV, Boyd S, et al. Circulating tumour DNA predicts response toanti-PD1 antibodies in metastatic melanoma [J]. Ann Oncol. 2017;28(5):1130–6.

108. Lee JH, Long GV, Menzies AM, et al. Association Between Circulating TumorDNA and Pseudoprogression in Patients With Metastatic Melanoma TreatedWith Anti–Programmed Cell Death 1 Antibodies [J]. JAMA Oncol. 2018;4(5):717–21.

109. Wang Z, Duan J, Cai S, et al. Assessment of blood tumor mutational burdenas a potential biomarker for immunotherapy in patients with non-small celllung Cancer with use of a next-generation sequencing Cancer gene panel[J]. JAMA Oncol. 2019;5(5):696–702.

110. Peters S, Cho BC, Reinmuth N, et al. Abstract CT074: Tumor mutationalburden (TMB) as a biomarker of survival in metastatic non-small cell lungcancer (mNSCLC): Blood and tissue TMB analysis from MYSTIC, a Phase IIIstudy of first-line durvalumab ± tremelimumab vs chemotherapy. Presentedat: American Association for Cancer Research Annual Meeting 2019. Atlanta,GA; 2019.

111. Chen G, Huang AC, Zhang W, et al. Exosomal PD-L1 contributes toimmunosuppression and is associated with anti-PD-1 response [J]. Nature.2018;560(7718):382–6.

112. Del Re M, Marconcini R, Pasquini G, et al. PD-L1 mRNA expression inplasma-derived exosomes is associated with response to anti-PD-1antibodies in melanoma and NSCLC [J]. Br J Cancer. 2018;118(6):820–4.

113. Gong B, Kiyotani K, Sakata S, et al. Secreted PD-L1 variants mediateresistance to PD-L1 blockade therapy in non-small cell lung cancer [J]. J ExpMed. 2019;216(4):982–1000.

114. Nakamura Y. Biomarkers for Immune Checkpoint Inhibitor-Mediated TumorResponse and Adverse Events [J]. Front Med (Lausanne). 2019;6:119.

115. Bedikian AY, Johnson MM, Warneke CL, et al. Prognostic factors thatdetermine the long-term survival of patients with unresectable metastaticmelanoma [J]. Cancer Investig. 2008;26(6):624–33.

116. Kelderman S, Heemskerk B, van Tinteren H, et al. Lactate dehydrogenase asa selection criterion for ipilimumab treatment in metastatic melanoma [J].Cancer Immunol Immunother. 2014;63(5):449–58.

117. Taniguchi Y, Tamiya A, Isa SI, et al. Predictive factors for poor progression-free survival in patients with non-small cell lung Cancer treated withNivolumab [J]. Anticancer Res. 2017;37(10):5857–62.

118. Weber JS, Tang H, Hippeli L, et al. Serum IL-6 and CRP as prognostic factorsin melanoma patients receiving single agent and combination checkpointinhibition [J]. J Clin Oncol. 2019;37:100.

119. Sosa A, Lopez Cadena E, Simon Olive C, et al. Clinical assessment ofimmune-related adverse events [J]. Ther Adv Med Oncol. 2018;10:1758835918764628.

120. Yuen KC, Liu LF, Gupta V, et al. High systemic and tumor-associated IL-8correlates with reduced clinical benefit of PD-L1 blockade [J]. Nat Med.2020;26(5):693–8.

121. Schalper KA, Carleton M, Zhou M. Elevated serum interleukin-8 is associatedwith enhanced intratumor neutrophils and reduced clinical benefit ofimmune-checkpoint inhibitors [J]. Nat Med. 2020;26(5):688–92.

122. Conforti F, Pala L, Bagnardi V, et al. Cancer immunotherapy efficacy andpatients' sex: a systematic review and meta-analysis [J]. Lancet Oncol. 2018;19(6):737–46.

123. Wu Y, Ju Q, Jia K, et al. Correlation between sex and efficacy of immunecheckpoint inhibitors (PD-1 and CTLA-4 inhibitors)[J]. Int J Cancer. 2018;143(1):45–51.

124. Fulop T, Larbi A, Kotb R, et al. Aging, immunity, and cancer [J]. Discov Med.2011;11(61):537–50.

125. Kugel CH 3rd, Douglass SM, Webster MR, et al. Age correlates with responseto anti-PD1, reflecting age-related differences in Intratumoral effector andregulatory T-cell populations [J]. Clin Cancer Res. 2018;24(21):5347–56.

126. Nishijima TF, Muss HB, Shachar SS, et al. Comparison of efficacy ofimmune checkpoint inhibitors (ICIs) between younger and olderpatients: a systematic review and meta-analysis [J]. Cancer Treat Rev.2016;45:30–7.

127. Felip E, Ardizzoni A, Ciuleanu T, et al. CheckMate 171: a phase 2 trial ofnivolumab in patients with previously treated advanced squamous non-small cell lung cancer, including ECOG PS 2 and elderly populations [J]. EurJ Cancer. 2020;127:160–72.


128. Dudnik E, Moskovitz M, Daher S, et al. Effectiveness and safety of nivolumabin advanced non-small cell lung cancer: the real-life data [J]. Lung Cancer.2018;126:217–23.

129. Murphy WJ, Longo DL. The surprisingly positive association betweenobesity and Cancer immunotherapy efficacy [J]. Jama. 2019;321(13):1247–8.

130. Wang Z, Aguilar EG, Luna JI, et al. Paradoxical effects of obesity on T cellfunction during tumor progression and PD-1 checkpoint blockade [J]. NatMed. 2019;25(1):141–51.

131. Baumann BC, Mitra N, Harton JG, et al. Comparative effectiveness of protonvs photon therapy as part of concurrent Chemoradiotherapy for locallyadvanced Cancer [J]. JAMA Oncol. 2019;6(2):237–46.

132. Routy B, Le Chatelier E. Gut microbiome influences efficacy of PD-1-basedimmunotherapy against epithelial tumors [J]. Science. 2018;359(6371):91–7.

133. Gopalakrishnan V, Spencer CN. Gut microbiome modulates response toanti-PD-1 immunotherapy in melanoma patients [J]. Science. 2018;359(6371):97–103.

134. Chaput N, Lepage P, Coutzac C, et al. Baseline gut microbiota predictsclinical response and colitis in metastatic melanoma patients treated withipilimumab[J]. Ann Oncol. 2019;30(12):2012.

135. Matson V, Fessler J, Bao R. The commensal microbiome is associated withanti-PD-1 efficacy in metastatic melanoma patients [J]. Science. 2018;359(6371):104–8.

136. Sivan A, Corrales L, Hubert N, et al. Commensal Bifidobacterium promotesantitumor immunity and facilitates anti-PD-L1 efficacy [J]. Science. 2015;350(6264):1084–9.

137. Parham P, Ohta T. Population biology of antigen presentation by MHC classI molecules [J]. Science. 1996;272(5258):67–74.

138. Havel JJ, Chowell D, Chan TA. The evolving landscape of biomarkers forcheckpoint inhibitor immunotherapy [J]. Nat Rev Cancer. 2019;19(3):133–50.

139. Chowell D, LGT M. Patient HLA class I genotype influences cancer responseto checkpoint blockade immunotherapy [J]. Science. 2018;359(6375):582–7.

140. Kvistborg P, Yewdell JW. Enhancing responses to cancer immunotherapy [J].Science. 2018;359(6375):516–7.

141. McGranahan N, Rosenthal R, Hiley CT, et al. Allele-Specific HLA Loss andImmune Escape in Lung Cancer Evolution[J]. Cell. 2017;171(6):1259–71.e1211.

142. Theodoraki MN, Yerneni SS, Hoffmann TK, et al. Clinical significance of PD-L1(+) Exosomes in plasma of head and neck Cancer patients [J]. Clin CancerRes. 2018;24(4):896–905.

143. Yang JC, Hughes M, Kammula U, et al. Ipilimumab (anti-CTLA4 antibody)causes regression of metastatic renal cell cancer associated with enteritisand hypophysitis [J]. J Immunother. 2007;30(8):825–30.

144. Teraoka S, Fujimoto D, Morimoto T, et al. Early Immune-Related AdverseEvents and Association with Outcome in Advanced Non–Small Cell LungCancer Patients Treated with Nivolumab: A Prospective Cohort Study [J]. JThorac Oncol. 2017;12(12):1798.

145. Judd J, Zibelman M, Handorf E, et al. Immune-related adverse events as abiomarker in non-melanoma patients treated with programmed cell death1 inhibitors [J]. Oncologist. 2017;22(10):1232–7.

146. Fujisawa Y, Yoshino K, Otsuka A, et al. Retrospective study of advancedmelanoma patients treated with ipilimumab after nivolumab: analysis of 60Japanese patients [J]. J Dermatol Sci. 2018;89(1):60–6.

147. Kim HI, Kim M, Lee SH, et al. Development of thyroid dysfunction isassociated with clinical response to PD-1 blockade treatment in patientswith advanced non-small cell lung cancer [J]. Oncoimmunology. 2017;7(1):e1375642.

148. Teulings HE, Limpens J, Jansen SN, et al. Vitiligo-like depigmentation inpatients with stage III-IV melanoma receiving immunotherapy and itsassociation with survival: a systematic review and meta-analysis [J]. J ClinOncol. 2015;33(7):773–81.

149. Freeman-Keller M, Kim Y, Cronin H, et al. Nivolumab in resected andUnresectable metastatic melanoma: characteristics of immune-relatedadverse events and association with outcomes [J]. Clin Cancer Res. 2016;22(4):886–94.

150. Rzepecki AK, Cheng H, McLellan BN. Cutaneous toxicity as a predictivebiomarker for clinical outcome in patients receiving anticancer therapy [J]. JAm Acad Dermatol. 2018;79(3):545–55.

151. Shukla SA, Bachireddy P, Schilling B, et al. Cancer-Germline AntigenExpression Discriminates Clinical Outcome to CTLA-4 Blockade [J]. Cell. 2018;173(3):624–33.e8.

152. Rodig SJ, Gusenleitner D. MHC proteins confer differential sensitivity toCTLA-4 and PD-1 blockade in untreated metastatic melanoma [J]. Sci TranslMed. 2018;10(450):eaar3342.

153. Rizvi H, Sanchez-Vega F, La K, et al. Molecular determinants of response toanti-programmed cell death (PD)-1 and anti-programmed death-ligand 1(PD-L1) blockade in patients with non-small-cell lung Cancer profiled withtargeted next-generation sequencing [J]. J Clin Oncol. 2018;36(7):633–41.

154. Althammer S, Tan TH, Spitzmüller A, et al. Automated image analysis ofNSCLC biopsies to predict response to anti-PD-L1 therapy [J]. J ImmunotherCancer. 2019;7(1):121.

155. Yu Y, Zeng D, Ou Q, et al. Association of Survival and Immune-RelatedBiomarkers with Immunotherapy in patients with non-small cell lungCancer: a meta-analysis and individual patient-level analysis [J]. JAMA NetwOpen. 2019;2(7):e196879.

156. Cristescu R, Mogg R, Ayers M, et al. Pan-tumor genomic biomarkers for PD-1checkpoint blockade-based immunotherapy[J]. Science. 2018;362(6411):eaar3593.

157. Balachandran VP, Łuksza M, Zhao JN, et al. Identification of uniqueneoantigen qualities in long-term survivors of pancreatic cancer [J]. Nature.2017;551(7681):512–6.

158. Sarkizova S, Hacohen N. How T cells spot tumour cells [J]. Nature. 2017;551(7681):444–6.

159. Jiang P, Gu S, Pan D, et al. Signatures of T cell dysfunction and exclusionpredict cancer immunotherapy response [J]. Nat Med. 2018;24(10):1550–8.

Publisher’s NoteSpringer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.

predictive biomarkers for cancer immunotherapy with …review open access predictive biomarkers for...

Documents